About

What I do now

I lead the infrastructure team in a regulated healthcare environment and own our production LLM gateway: EU-resident model endpoints with a cross-provider fallback, and scoped per-service keys and budgets. Outside work, I am building a GPU inference lab on AWS EKS, run in short burst sessions, to measure what self-hosted LLM serving actually costs on spot A10Gs; the first GPU session is still to come. Earlier side projects (Kortex, an LLM API gateway operator, and an AI FinOps prototype) taught me where to build and where to adopt: today I would use the Gateway API Inference Extension for in-cluster model routing rather than my own router. The thread connecting all of it: making AI infrastructure observable, governable, and cost-transparent as a platform concern.

How I got here

Growing up in Nigeria, frequent power outages weren't just inconveniences - they were lessons in resilience. That shaped how I build: platforms designed for failure, with redundancy, fast recovery, and alerts that point at causes.

I started as an SAP technical administrator at the Nigerian National Petroleum Corporation, then moved through on-premise infrastructure, fintech on GCP, eCommerce at scale on AWS, defence-intelligence data platforms on Azure, and now healthcare. Across two platform programmes I delivered €900K+ in annual savings (see Work), and I have mentored 20+ engineers from their first Terraform module to owning platforms of their own.

Toolbox

Cloud and orchestration

AWS (EKS), GCP, Azure (AKS), Kubernetes, Karpenter, KEDA

AI infrastructure

LiteLLM, vLLM, KServe, Vertex AI, Amazon Bedrock

Infrastructure as code and GitOps

Terraform, Terragrunt, Pulumi, Crossplane, Helm, ArgoCD, GitHub Actions

Observability

Prometheus, Grafana, Loki, Tempo, OpenTelemetry, OpenSearch

Security and policy

Kyverno, cosign, IRSA and workload identity, External Secrets

Languages

Go, Python, TypeScript, Bash, HCL

How I work

• Platform as product

  Internal platforms have users. Developer experience is a requirement, not a nice-to-have.

• Everything in Git

  If it is not in version control, it does not exist. Changes go through review and the pipeline.

• Measure, then claim

  SLOs over vanity metrics, attributed cost over total spend, and measured results over design targets.

• Change your mind on evidence

  Write the decision down with its alternatives, and reverse it when the facts change.

Certifications

• Certified Kubernetes Administrator (CNCF)
• Certified Kubernetes Application Developer (CNCF)
• Kubernetes & Cloud Native Associate (CNCF)
• Kubernetes & Cloud Native Security Associate (CNCF)
• AWS Certified Solutions Architect Associate (Amazon)
• AWS Certified Cloud Practitioner (Amazon)
• Microsoft Certified Azure Fundamentals (Microsoft)
• HashiCorp Certified Terraform Associate (HashiCorp)
• Codefresh GitOps Certified Fundamentals (Codefresh)
• Codefresh GitOps Certified Scale (Codefresh)
• Aviatrix Certified Engineer - Multi-Cloud Network Associate (Aviatrix)